Microsoft India & South Asia president Puneet Chandok, in an interview with Poulomi Chatterjee, talks about the pace of AI adoption in India, the changing nature of tech jobs and how enterprises can extract returns from agentic AI. Excerpts:
Q. Reports say India is enthusiastic about AI adoption but has the largest number of projects stuck in pilot. Is that changing?
I disagree with that perception. India is way past that stage. We had a lot of proof of concepts (POC) during the experimentation phase, rightly so, but we are now fully in the production phase. Air India is anchoring one of the largest customer-experience reinventions globally. Infosys, TCS and Wipro have each deployed 100,000 Copilots, making AI part of the fabric of Indian IT services. Apollo Hospitals is redesigning critical care, while Adani Group and Axis Bank are among enterprises pushing AI into production. We have gone through what I call “pilot hell”, and the work is now quite advanced.
Q. Will AI-led hiring increase as productivity gains create more work?
That is India’s advantage. We are already seeing new job roles such as forward-deployed engineers (FDE), enterprise ontology specialists, translators and agent orchestrators — roles that were not even in my vocabulary six months ago. At the same time, the nature of existing jobs is changing. Roles will get unbundled, with each one comprising a portfolio of tasks. But this is still a transition. That is why Microsoft has a skilling commitment covering 20 million people in India and 2 million teachers through its Elevate programme.
Q. What makes someone a frontier professional? Is that different from a forward-deployed engineer?
The intensity of AI usage is the starting point. AI has to become part of your everyday work as ambient intelligence. Frontier professionals use AI for reasoning, not just transactional tasks. But that is table stakes. They should be able to build and orchestrate AI agents for multi-step processes and draw intelligence from them. They should also be able to reimagine existing workflows. Most importantly, they need to retain the human end of the bargain. AI can do a lot, but judgment is still absent. Critical thinking, quality control, interrogating AI outputs and deciding where AI should — and should not — be used are vital. FDEs and frontier professionals are separate terms. If you are an FDE, you will be a frontier professional, but it does not work the other way.
Q. Does this raise the bar for hiring freshers?
We are looking for new requirements, and this creates a barbell in hiring. On one side is the ability to orchestrate agents, reimagine processes and understand domains. On the other are human skills such as judgment, sense-making, curiosity, empathy and the ability to bring energy and clarity to work. The middle layers of execution, orchestration and data crunching will increasingly be commoditised.
Q. Is agentic AI delivering ROI beyond productivity gains?
It is already happening across sectors. We have seen examples such as Persistent Systems’ AI assistant PiAssist, which has reinvented customer experiences, and Wipro’s iSkill programme, which is scaling employee experiences with better outcomes. These are not just productivity improvements; they are entirely new capabilities. The next step is to scale them across enterprises and public-sector institutions.
Q. As AI agents become more pervasive, how should enterprises control costs?
It is not about spending less on tokens but spending intentionally. Enterprises have to consider token costs, quality of outputs, speed, latency and regulatory compliance. The answer is an open, heterogeneous system with multiple AI models. Microsoft Foundry has 11,000 models. Enterprises need to use frontier models judiciously — choosing the right model for the right use case — and build their own agent harness around it, with trust at the centre.